A quick question on tris/meshes and performance

Hey guys, I am currenty working on a racing game. I have the first track built, collider wise, anyway. I haven’t populated it, textured it or anything. Anyway, it is working brilliantly but by the time it’s finished it will be pretty high poly. My question is, before I get started on the visual aspects, would having two meshes, one for collision (high poly) and one for rendering (lower poly) be better for performance than a single high poly mesh? I’m unsure as to how Unity handles this stuff and don’t want to spend a month on an awesome looking environment to find it runs at 3fps. lol. Basically, I can drop the polycount at least in half and it will still look good, just wouldn’t be good enough for smooth collision.

For games with high poly graphics, you do generally use physics proxies to approximate the geometry to make physics calculations fewer and less prone to bugginess.

To my understanding, Unity works best with convex meshes as colliders. So if you’re doing something with lots of concave parts, you’ll get best performance and reliability by breaking it into convex parts and using mesh colliders, or just approximating tricky shapes with the primitive box / capsule / sphere colliders, leaving mesh colliders for simple geometry like ground or walls.

In general though, the primitive colliders have far better performance than mesh colliders.